Minor Comment: You seemed to compare the "Dhrystone" figures, but that's an Integer-Benchmark anyways. The relevant figure to check the FPU performance is the "MFlops" Benchmark figure. It was still equal in both your tests, at 0.43 MFlops. Which is exactly what you can expect from the 68882 running at 14 MHz.
Yep, a direct install will clock the FPU at half the oscillator clock, just like the onboard EC020. But you can lift pin 11 of the FPU (or the PLCC connector, like I did) and jump it directly to E122R (just above the oscillator) and get the full 28MHz clock.
Makes sense, as the CPU Clock signal from the oscillator is tied to both the 68881/68882 and the 68020, so yes it will pull the A1200 motherboard Clock. The circuitry for ascertaining the primary CPU between the Onboard and Trapdoor, is a simple circuit that ties the 68020 in to a never ending wait for Bus control, while the trapdoor receives an unimpeded Bus Grant signal. Same for the 600, with the 68000 on the motherboard and the clip on CPU/FPU boards, force the 68000 in to a never ending wait for Bus control. At power up, the onboard CPU sends out a _BR (Request Bus control), then the CPU sits and waits for a _BG (Bus request granted), but what you do is tie the _BG pin to ground via a 1k resistor, and the chip just sits there waiting. Then on the expansion board, you tie VCC (with 1k Resistor) to the new board's CPU pin (_BG) so it gets permission over the system bus, and off it goes. Where you are going to run in to problems is the Expansion having a 68882, as well as one soldered (or by socket) on the motherboard, then you will have 2 devices trying to take up the same address space, likely cancelling each other out.. But having said which, the 68882 is a rather poor floating point unit. imgur.com/a/fO2m5xA
The FPU certainly speeds up rendering. You can really see the difference in WinUAE if you toggle FPU on and off and hit the ol' LightWave render button (make sure you're running the FP version of LW3D, of course).
I have an FPU slot on my Blizzard A1230 MKII and it accepts a 50 MHz FPU and runs at the same speed as the CPU. If possible, its best to get an FPU on the same board and have them running at the same speed. Its awesome to know that the A1200 can have an FPU mounted onto the motherboard, I've always wondered if the traces actually went where they could be of use. As I had an accelerator with an FPU anyway, I never tried to put one on the motherboard. The IDE interface on the A1200 does support 2 drives (4 if you get a splitter) and I found that a modern laptop IDE SSD will fit nicely into the hard drive mount and is thinner than the Amiga's original hard drive. I then used double sided tape on both sides of some plastic to insulate the under side of the IDE to CF adapter and stuck it to the top of the SSD. I then made a custom cable with 3 mini IDE sockets and secured the drives to the mount. My main Amiga 1200 setup HD floppy drive for 880 K and 1.44 MB disks (ROMs need upgrading to use HD drive) 3 x External floppy drives 880 K 8 GB IDE SSD - 2 x 4 GB partitions 8 GB IDE to CF adapter - 2 x 4 GB partitions Blizzard A1230 MKII @ 50 MHz + FPU 128 MB RAM Blizzard SCSI interface card 128 MB RAM External SCSI HDD x 2 @ 60 GB - 4 GB partitions External CD-RW x 2 External SCSI backup tape drive SCSI scanner Total RAM 256 MB + 2 MB on board
I've not tested between motherboard FPU and one on expansion, so I can't comment. I would assume it would be like a lot of hardware, where it is almost always better to run directly on the motherboard (obviously don't try to run 2x FPU's at the same time). If you do this mod, please let us know in the comments which gives the best results. Yes, it's absolutely worth putting an SSD in the A1200. I put a 128GB M.2 SSD in mine without any problem, I made a video about that upgrade too.
If you remove the bottom of the socket, is the connectors along the midpoints of each side still being able to make a good contact over time? Because all those springy contacts would want to bulge the socket outwards methinks, and with only the solder points on the board holding things together it might not be enough. Maybe. I dunno! :D Also, PLCC sockets tend to be kind of brittle and easy to crack; for longevity I'd rather keep the bottom of the socket in place I'd think... :)
yes, I was thinking this was to be put in a stock A1200, but it clearly has an 030 CPU that is faster than an A1200 one, but there does not seem to be any mention about the CPU having been replaced or an expansion board. I am pretty sure the FPU did not make the 020 seem like an 030 and just be faster after putting the FPU in?
@@perinoid yes, so it does not make much sense why the video is all about putting an FPU into an A1200 and then having some expansion card in that A1200 as well. Seems more interesting to test it only with the built in CPU and then maybe also compare it to tests with an expansion card. Also would it work with something like the ACA1221 cards? Seems plausible.
@@KasperSOlesen Terrible fire card TF1230 is without FPU - so it make sense to add FPU to the motherboard . FPU works synchronous with CPU - so it base is 14MHZ - but if you pull up the 11 pin from FPU and connect it directly to the oscillator on board then you have a FPU on 28Mhz.
@@flowcomp sure, it makes sense to want 030 and FPU. But the video mentioned nothing about "030 expanded Amiga 1200" and fails to even mention directly that it needs an expansion to use this, you have to figure it out by reading the benchmark screen.
Minor Comment: You seemed to compare the "Dhrystone" figures, but that's an Integer-Benchmark anyways. The relevant figure to check the FPU performance is the "MFlops" Benchmark figure.
It was still equal in both your tests, at 0.43 MFlops. Which is exactly what you can expect from the 68882 running at 14 MHz.
Yep, a direct install will clock the FPU at half the oscillator clock, just like the onboard EC020. But you can lift pin 11 of the FPU (or the PLCC connector, like I did) and jump it directly to E122R (just above the oscillator) and get the full 28MHz clock.
Great info there, Sir! I will be trying this out for sure. Thank you!
Makes sense, as the CPU Clock signal from the oscillator is tied to both the 68881/68882 and the 68020, so yes it will pull the A1200 motherboard Clock.
The circuitry for ascertaining the primary CPU between the Onboard and Trapdoor, is a simple circuit that ties the 68020 in to a never ending wait for Bus control, while the trapdoor receives an unimpeded Bus Grant signal. Same for the 600, with the 68000 on the motherboard and the clip on CPU/FPU boards, force the 68000 in to a never ending wait for Bus control.
At power up, the onboard CPU sends out a _BR (Request Bus control), then the CPU sits and waits for a _BG (Bus request granted), but what you do is tie the _BG pin to ground via a 1k resistor, and the chip just sits there waiting. Then on the expansion board, you tie VCC (with 1k Resistor) to the new board's CPU pin (_BG) so it gets permission over the system bus, and off it goes.
Where you are going to run in to problems is the Expansion having a 68882, as well as one soldered (or by socket) on the motherboard, then you will have 2 devices trying to take up the same address space, likely cancelling each other out.. But having said which, the 68882 is a rather poor floating point unit.
imgur.com/a/fO2m5xA
The FPU certainly speeds up rendering. You can really see the difference in WinUAE if you toggle FPU on and off and hit the ol' LightWave render button (make sure you're running the FP version of LW3D, of course).
I have an FPU slot on my Blizzard A1230 MKII and it accepts a 50 MHz FPU and runs at the same speed as the CPU. If possible, its best to get an FPU on the same board and have them running at the same speed.
Its awesome to know that the A1200 can have an FPU mounted onto the motherboard, I've always wondered if the traces actually went where they could be of use. As I had an accelerator with an FPU anyway, I never tried to put one on the motherboard.
The IDE interface on the A1200 does support 2 drives (4 if you get a splitter) and I found that a modern laptop IDE SSD will fit nicely into the hard drive mount and is thinner than the Amiga's original hard drive. I then used double sided tape on both sides of some plastic to insulate the under side of the IDE to CF adapter and stuck it to the top of the SSD. I then made a custom cable with 3 mini IDE sockets and secured the drives to the mount.
My main Amiga 1200 setup
HD floppy drive for 880 K and 1.44 MB disks (ROMs need upgrading to use HD drive)
3 x External floppy drives 880 K
8 GB IDE SSD - 2 x 4 GB partitions
8 GB IDE to CF adapter - 2 x 4 GB partitions
Blizzard A1230 MKII @ 50 MHz + FPU
128 MB RAM
Blizzard SCSI interface card
128 MB RAM
External SCSI HDD x 2 @ 60 GB - 4 GB partitions
External CD-RW x 2
External SCSI backup tape drive
SCSI scanner
Total RAM 256 MB + 2 MB on board
I've not tested between motherboard FPU and one on expansion, so I can't comment. I would assume it would be like a lot of hardware, where it is almost always better to run directly on the motherboard (obviously don't try to run 2x FPU's at the same time). If you do this mod, please let us know in the comments which gives the best results.
Yes, it's absolutely worth putting an SSD in the A1200. I put a 128GB M.2 SSD in mine without any problem, I made a video about that upgrade too.
To really test the FPU, you need AIBB65 and to enable 020/FPU option.
The "golden" text buttons, are the FPU tests.
You don't need to knock the bottom out of your socket to solder it, just use a 0.3mm soldering iron tip.
If you remove the bottom of the socket, is the connectors along the midpoints of each side still being able to make a good contact over time? Because all those springy contacts would want to bulge the socket outwards methinks, and with only the solder points on the board holding things together it might not be enough. Maybe. I dunno! :D
Also, PLCC sockets tend to be kind of brittle and easy to crack; for longevity I'd rather keep the bottom of the socket in place I'd think... :)
The more important question is: how does it behave with only an on-board 68EC020?
yes, I was thinking this was to be put in a stock A1200, but it clearly has an 030 CPU that is faster than an A1200 one, but there does not seem to be any mention about the CPU having been replaced or an expansion board.
I am pretty sure the FPU did not make the 020 seem like an 030 and just be faster after putting the FPU in?
@@KasperSOlesen Of course not. The raw benchmark speed is equal to the one of a 1230+FPU accelerator (~17x of the basic A600).
@@perinoid yes, so it does not make much sense why the video is all about putting an FPU into an A1200 and then having some expansion card in that A1200 as well. Seems more interesting to test it only with the built in CPU and then maybe also compare it to tests with an expansion card.
Also would it work with something like the ACA1221 cards? Seems plausible.
@@KasperSOlesen Terrible fire card TF1230 is without FPU - so it make sense to add FPU to the motherboard . FPU works synchronous with CPU - so it base is 14MHZ - but if you pull up the 11 pin from FPU and connect it directly to the oscillator on board then you have a FPU on 28Mhz.
@@flowcomp sure, it makes sense to want 030 and FPU. But the video mentioned nothing about "030 expanded Amiga 1200" and fails to even mention directly that it needs an expansion to use this, you have to figure it out by reading the benchmark screen.